[EM] IIAC. Juho: Census re-districting instead of PR for allocating seats to districts.

Juho Laatu juho4880 at yahoo.co.uk
Sat Jun 16 06:35:10 PDT 2012


On 16.6.2012, at 5.33, Michael Ossipoff wrote:

> You said:
> 
> Can you give an example where there is a difference between those
> methods?

> (Sainte-Lague and Largest-Remainder)
> 
> Yes, in the sources that I've already referred you to. Do you really
> need me to look it up for you?

No need. I think I still understand what we are talking about. Just tell which property is the one that makes LR unacceptable to you when compared to SL.

> You spoke above of Largest-Remainder's advantages. Name one.Explaining
> the method briefly to someone who isn't interested in PR?

I gave one explanation few mails back. (http://lists.electorama.com/pipermail/election-methods-electorama.com/2012-June/030539.html, see the beginning of the mail) The idea was roughly that if we could give fractional seats we would probably use the LR formula (Hare quota). That gives us one possible natural measure of deviation from that ideal, and that can be used to speak in favour of LR. I guess that's pretty much why LR has been proposed in the first place.

> Starting with a Sainte-Lague seat allocation to districts, take a seat
> from one district and give it to another. That will always result in a
> greater difference between those two districts' seats per person.

Following this property, with proportions 3-3-1 SL allocates two first seats to the two large groups, and considers all three groups to have equal rigth to the third seat. LR says that the third seat should go to the smallest group. From LR point of view, giving the third seat to the smallest group means that the largest groups will each have 0.667 voters that are not represented (since each representative should ideally represent 2.333 voters), and the smallest group is overrepresented worth 1.333 ghost voters. The overrepresentations are thus -0.667, -0.667 and 1.333 voters. If one gives the third seat to the first group, the overrepresentations will be 1.667, -0.667 and -1.000 voters, i.e. worse. Change of allocation from 1-1-1 to 2-1-0 would thus change the overall need to move people/voters from one district to another (= "sum of errors in districting") from 2.667 voters to 3.333 voters  (the target of the changes is to make the representation equal). That is one possible argument to follow the LR philosophy. LR thus focuses more on the number of voters whose rights are violated while SL focuses on the proportion (proportion changes in a small group means less people/voters than in a large group).

(Btw, was this the key property that makes LR unacceptable to you, when compared to SL?)

>> Is Sainte-Laguë's seat allocation 1-1-1 with proportions 57-23-20 ok? Note that this introduces an incentive to split a minority district (43) in two parts (23-20) to get majority of the seats. (may lead to gerrymandering)

> ... if you're claiming that there are examples of a
> splitting-incentive in Sainte-Lague, but not in Largest-Remainder,
> then you should say so.

No intention to say so. I read your answer so that the presented propoerty is a problem, but one can fight against gerrymandering e.g. by using automatic districting. The purpose of that question was just to check that you don't claim that SL would always give optimal results or better results than other methods.

Juho




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20120616/a9be5e8f/attachment-0003.htm>


More information about the Election-Methods mailing list